In 2024, the Korea Institute of Ocean Science and Technology (KIOST) and the Korea Institute of Science and Technology (KIST) conducted joint comprehensive marine atmospheric observations onboard the R/V ISABU with a gross tonnage of 5894 tons, a length of 99.8 m, and a width of 18.0 m (Fig. S1 in the Supplement). The primary objective of this campaign was to jointly investigate interactions among marine biology, marine atmosphere, cloud formation, and climate change. This was achieved through simultaneous measurements of marine aerosols, CCN, ice-nucleating particles (INP), and gaseous species, as well as in situ observations of atmospheric photochemical processes.

The transit voyage departed from Jangmok Port (34.993° N, 128.676° E) in Geoje, Republic of Korea, on 28 March and arrived in Port Louis (20.152° S, 57.494° E), Mauritius, on 17 April, passing through the East China Sea, the South China Sea, the Strait of Malacca, and the Indian Ocean. The detailed cruise route and schedule are presented in Fig. 1 and Table S1 in the Supplement. The fact that measurements were conducted across diverse sea areas demonstrates the strength of this campaign, enabling consistent comparisons of regional characteristics.

During the transit voyage of the R/V ISABU, various instruments were installed to measure both particulate and gaseous components in the atmosphere, as shown in Fig. S1. The observation room was located on the uppermost F-deck (approximately 20 m above sea level) to minimize the influences of sea-salt particles and ammonia emitted by the vessel. All aerosol observation instruments were connected via a conductive tube to reduce losses. To limit the effects of high ambient humidity, a diffusion dryer was installed upstream of each instrument inlet. Nevertheless, due to the substantial temperature difference between the room and the outside, condensation occasionally occurred inside the tube, requiring periodic checks and the removal of water.

Among the various instruments, a scanning mobility particle sizer (SMPS; model 3938L50, TSI Inc., USA) and a cloud condensation nuclei counter (CCNC; model CCN-200, Droplet Measurement Technologies, USA) were used in this study. SMPS, which consists of an electrostatic classifier (Model 3082, TSI Inc., USA) and a condensation particle counter (CPC; model 3750, TSI Inc., USA), measured the aerosol size distribution every 3 min. The measured aerosol size range was 10.6–478.3 nm, divided into 107 bins. However, occasional voltage instabilities occurred at the smallest size bins, so the 10.6–14.6 nm range was excluded when calculating NCN. Accordingly, NCN in this study refers to the total number concentration of aerosols larger than 15 nm. In addition, to investigate the size-resolved variations of aerosols, we categorized the SMPS size range into three modes: nucleation mode (15–25 nm), Aitken mode (25–100 nm), and accumulation mode (100–478.3 nm).

CCNC measured NCCN at a given supersaturation every second. Although the CCNC model used in this study consisted of two columns, only one was operated during the campaign due to a flow issue caused by a Nafion tube in the other. CCNC was set to run at supersaturations ranging from 0.2 % to 1.0 %, with increments of 0.2 %. The measurement durations were 12 min for 0.2 % SS and 5 min for the 0.4 %–1.0 % SS. Thus, the nominal total measurement cycle of CCNC was 32 min. However, the instrument software did not initiate measurements at the next supersaturation immediately after completing the previous one. That is, the measurement timer began only after the temperature gradient between the top and bottom of the column reached the set value for the next supersaturation. As a result, the actual total measurement cycle during the cruise was approximately 50 min.

Furthermore, the time required for temperature stabilization varied even at the same supersaturation. Therefore, for efficient data processing, we uniformly excluded the first 15 min of data at 0.2 % SS and the first 3 min of data at the other supersaturations, regardless of the actual measurement start time. The relatively longer exclusion period for 0.2 % SS was due to the substantial change in supersaturation from 1.0 % to 0.2 %. Unlike other changes, which took about 90 s to stabilize the temperature, this specific change required about 10 min. To ensure accurate measurements, we performed CCNC flow calibration and supersaturation calibration using (NH4)2SO4 before and after the campaign.

To ensure reliable measurement data and enable consistent comparisons across different sea areas, we conducted data quality control following the procedure described below.

First, since the ship plume stack was located behind the observation room, measurements might be contaminated by the vessel's exhaust when the wind blew rapidly from the stern (Fig. 2a). Therefore, data were excluded when the relative wind direction was between 150 and 270°, considering the relative positions of the inlet and plume stack. Fortunately, the wind blew from the bow during most of the cruise (Fig. 2b), and contamination by exhaust was minimal at 2.8 %.

Second, temporary contamination could also occur due to ship maintenance activities such as painting and cleaning, cooking on board, or exhaust plumes from nearby ships (particularly in the Strait of Malacca). Such contamination was removed through a spike check. Using a 24 h moving window, any data point that deviated from the median by more than 3 times the scaled Median Absolute Deviation (MAD) was considered an outlier and excluded from the dataset.

Lastly, we corrected the SMPS data. Due to an unidentified issue, likely related to relative humidity, NCN occasionally exhibited a nearly order-of-magnitude change while maintaining a consistent size distribution (Fig. S2 in the Supplement). If these variations reflected real atmospheric changes, they should be clearly evident in the non-refractory PM₁ data measured by the Aerosol Mass Spectrometer. However, since no such signal was observed (Fig. S3 in the Supplement), the SMPS data are presumed to be anomalous. Consequently, anomalous SMPS data were identified and corrected through comparison with the aerosol size distribution measured by the Aerodynamic Particle Sizer (APS). The corrected data were then validated against PM_2.5 data obtained by the beta attenuation method before being used for this study (Fig. S4 in the Supplement). Details of the correction procedure are provided in the Supplement (Sect. S1).

After applying these three procedures (including the exclusion mentioned in Sect. 2.1.2 for the CCNC), 86.8 % of the SMPS data and 36.6 % of the CCNC data were used for analysis, which correspond to approximately 17.4 and 7.33 d of data, respectively.

Merged SMPS–APS size distributions (15–3000 nm) were fitted with multiple lognormal modes to characterize the aerosol modal structure, where each mode was expressed as follows: 1Ni(Dp)=Nt,i2πlog⁡10σg,i×exp⁡-(log⁡10Dp-log⁡10Dpg,i)22(log⁡10σg,i)2, where Dp denotes the aerosol diameter. Nt,i, Dg,i, and σg,i are the total number concentrations, the geometric mean diameter, and the geometric standard deviation of mode i, respectively.

To better reproduce the broad large-particle shoulder, following the concept of Modini et al. (2015), the coarse mode was first fitted to the upper part of the size distribution (1000–2500 nm) using a constrained single lognormal mode, with bounds of Nt=0–20 cm-3, Dpg=500–1800 nm, and σg=1.5–3.0. These bounds were selected to allow a single broad mode to represent the observed coarse mode shoulder while minimizing interference from the dominant submicron modes. In addition, while the upper limit of the merged size distribution is 3000 nm, the coarse mode fitting was only performed up to 2500 nm to avoid overfitting in the 2500–3000 nm range, where number concentrations were negligible.

The fitted coarse mode was then subtracted, and the residual submicron distribution (15–1000 nm) was fitted with multiple lognormal modes. The maximum number of submicron modes was limited to three to preserve physically interpretable mode structures. The optimal number of modes was determined sequentially: when the addition of one extra mode decreased the RMSE by less than 10 %, the previous number of modes was selected as optimal. The parameter bounds for the submicron modes were basically Dpg=15–25 nm and σg=1.0–1.5 for the nucleation mode, Dpg=25–100 nm and σg=1.0–2.0 for the Aitken mode, and Dpg=100–500 nm and σg=1.0–2.0 for accumulation mode, with Nt=0–10 000 cm-3 for all three.

To calculate the critical diameter (DC) at each supersaturation, we assumed that the aerosols were internally mixed and that larger aerosols would be preferentially activated due to smaller curvature effects. First, NCCN, originally recorded at 1 s intervals, were averaged to match the 3 min sampling interval of the SMPS. Then, NCN from the largest to the smallest diameter bins of the observed size distribution were cumulatively summed. As explained in Eq. (2), the diameter of the bin where the cumulative NCN first exceeded the NCCN was taken as DC. 2NCCN<∫DC478.3nmNCN(Dp)dDp.

The hygroscopicity parameter (κ) represents the relationship between dry diameter and CCN activity and was calculated according to Petters and Kreidenweis (2007): 3κ=4A327Dc3ln⁡2Sc,A=4σs/aMwRTρw, where σs/a is the surface tension of the solution/air interface, and in this study, it was assumed to be 0.072 Jm-2, corresponding to the surface tension of pure water. Mw=0.018 kgmol-1 and ρw=997 kgm-3 (at 273.15 K) are the molar mass and density of water, respectively. R=8.3145 Jmol-1K-1 is the universal gas constant. T is the temperature; in this study, the measured temperature was used. Sc refers to the supersaturation used in the calculation of the corresponding Dc.

To identify the origin and transport pathway of air masses during the cruise, back trajectories at 100 m altitude were calculated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Stein et al., 2015). For these calculations, the coordinates of the research vessel at each hour were used as the starting location. The total run time was set to 72 h by default, but was extended to 120 h when additional analysis was required. The meteorological input for HYSPLIT was sourced from the Global Data Assimilation System (GDAS) model with a 1°×1° horizontal resolution.

To classify the origins of air masses over each sea area, k-means clustering analysis was conducted using the collected back-trajectory dataset. Fundamentally, clustering analysis requires all trajectories to share an identical starting location. However, because the starting locations of the acquired trajectories differed, the three-step preprocessing procedure (translation, vectorization, and standardization) was applied, as explained in detail in the Supplement (Sect. S2). The k-means clustering performed after this preprocessing had a limitation that it did not fully retain the actual geographic information of the trajectories. Therefore, while the k-means clustering results served as the primary basis, the geographic information of the original trajectories was additionally taken into account to finalize the classification.

Figure 3 presents the time series of aerosol size distribution (Fig. 3a), NCN and NCCN (Fig. 3b), temperature and relative humidity (RH) (Fig. 3c), true wind speed and direction (Fig. 3d), and precipitation and pressure (Fig. 3e) during the transit voyage of the R/V ISABU. Since the local time zone was adjusted five times in total (from UTC+9 to UTC+4), the Coordinated Universal Time (UTC) is shown in black along the bottom x axis, and the Local Time (LT) is shown in gray along the top x axis. The defined areas for each sea region are presented in Fig. 1.

NCN ranged from 11 to 9171 cm-3, with an average of 695 ± 889 cm-3, and exhibited a distinct contrast between the three seas–the East China Sea, the South China Sea, and the Strait of Malacca–and the Indian Ocean, with longitude 90° E serving as a dividing boundary (Fig. 4a). In terms of aerosol size, the particle size range was 50–300 nm until 9 April (i.e., at 90° E), excluding the first day of the cruise (28 March, LT), and thereafter it showed a slightly smaller range of 20–200 nm. The Aitken and accumulation modes also showed apparent differences across 90° E (Fig. 4d and e), indicating that the Indian Ocean is characterized by conditions close to those of a background atmosphere.

Meanwhile, NCCN ranged from 1 to 2407 cm-3 at 0.2 % SS and from 4 to 2670 cm-3 at 0.6 % SS, with averages of 146 ± 232 cm-3 and 323 ± 348 cm-3, respectively, and showed a clear contrast across 90° E (Figs. 5a, b and S5a in the Supplement). Although CCN covered a somewhat narrower concentration range than the total aerosols, the two concentrations exhibited remarkably similar spatiotemporal variability throughout the entire period. However, the NCCN/NCN ratios displayed the opposite trend, with notably higher ratios in the Indian Ocean than in other regions, particularly at higher supersaturation (Figs. 5c, d and S5b).

In the South China Sea, between 16:30 on 1 April and 02:30 on 2 April (UTC), a noticeable increase in both NCN and NCCN of nearly an order of magnitude was observed in the western part of Luzon Island, Philippines (near 15.5° N, 116.8° E), accompanied by high NCCN/NCN ratios. Notably, NCCN during this period reached its maximum value for the entire cruise. A crucial point is that this period was the only instance in which the NCCN at 0.2 % SS, where only relatively large aerosols can activate, were similar to NCCN at the other supersaturations (Fig. 3b). In fact, this enhancement was primarily driven by the accumulation mode (Fig. 4e), particularly by particles around 200 nm in diameter, consistent with the distribution of geometric mean diameter (Dg) presented in Fig. 4b. A detailed analysis of this phenomenon will be provided in Sect. 3.1.2.

Localized variation was observed within the Strait of Malacca. NCN were higher upon entering the strait from the South China Sea than upon exiting into the Indian Ocean (2153 ± 1257 cm-3 when entering vs. 901 ± 212 cm-3 when exiting). This pattern was also evident in the spatial distributions of the three aerosol modes. When combined with the aerosol size distribution, the contrast was particularly pronounced in the Aitken mode, suggesting influence from the ship exhaust. However, the distribution of the geometric mean diameter was larger when exiting the strait (85.6 ± 17.2 nm when entering vs. 113.5 ± 17.7 nm when exiting). This is because the ratio of Aitken mode to accumulation mode was lower upon exiting the strait (1.32 ± 0.80 when entering vs. 0.49 ± 0.20 when exiting). NCCN at 0.2 % SS exhibited a relatively uniform concentration within the strait, whereas NCCN at the other supersaturations were higher upon entering the strait (at 0.6 % SS, 753 ± 404 cm-3 when entering vs. 332 ± 57 cm-3 when exiting), consistent with the pattern of NCN. As a result, the NCCN/NCN ratios did not show substantial variations within the strait.

Over the Indian Ocean, on 9 April (UTC), NCN and NCCN decreased sharply, reaching their minimum values (NCN: 11 cm-3; NCCN at 0.2 % and 0.6 % SS: 1 and 3 cm-3, respectively) during the cruise. One contributing factor to this decrease was the wet deposition associated with intense precipitation. Beginning around 22:00 on 8 April (UTC), heavy rainfall occurred for about 2 h, with a peak precipitation rate of 60 mmh-1, likely scavenging a substantial fraction of atmospheric particles. Meanwhile, in the Indian Ocean, NCN and NCCN (at 0.2 % and 0.6 % SS) during non-precipitating periods were 150 ± 230, 30 ± 34, and 67 ± 60 cm-3, respectively, which were higher than the average values during the precipitating periods (130 ± 209, 11 ± 12, and 37 ± 36 cm-3). However, the difference between NCN and NCCN was similar across both periods. Separately, a reduction in number concentrations due to intense precipitation was also observed on the first day of the cruise over the Southern Sea area of Korea.

During the period from 16:30 on 1 April to 02:30 on 2 April (UTC), high NCN and NCCN were observed as the R/V ISABU passed west of Luzon Island, Philippines. Compared with the periods before and after this event, NCN increased by approximately 1844 cm-3 (186 %), and NCCN at 0.2 % and 0.6 % SS increased by about 1167 cm-3 (398 %) and 1037 cm-3 (151 %), respectively.

The source of this high concentration was apparently the Taal volcano, located in the southern part of Luzon (14.010° N, 120.998° E). According to the Volcano Bulletins provided by the Philippine Institute of Volcanology and Seismology (PHIVOLCS), this active volcano exhibited intermittent eruptive activity throughout 2024. Notably, the highest SO2 flux of 18 639 tons in 2024 was observed on 28 March, just before the high-concentration period (Fig. S6 in the Supplement). Another substantial flux of over 10 000 tons was recorded two days later, on 30 March. To assess whether emissions from the Taal volcano could have influenced the South China Sea, a back trajectory analysis was conducted. The starting altitudes for the trajectories were set to 500, 750, and 1000 m, accounting for the volcano's elevation (311 m) and the reported plume heights (900–1200 m).

Figure 6a shows the air mass back trajectories originating from three starting heights above the Taal volcano. In all cases, the trajectories passed through the region where high NCCN at 0.2 % SS were observed. However, if an air mass passes over a specific region above the boundary layer (BL), it may not influence the surface. Therefore, ERA5 reanalysis data were used to compare the BL height at the location of the R/V ISABU with the heights of the trajectories (Fig. 6b). The red dots indicate the times and heights at which each trajectory was closest to the cruise route. Many of these points lie below the BL height (indicated by the blue line) during the high-concentration period (shaded in green). This confirms that the conditions were conducive to volcanic influence reaching the near-surface marine boundary layer.

SO2 measured by the Air Quality Monitoring System (AQMS; Thermo Fisher Scientific Inc., USA) installed on the R/V ISABU also showed significantly elevated concentrations in the western part of Luzon (Fig. S7 in the Supplement), providing strong evidence that the Taal volcano was the source responsible for the high NCN and NCCN. Importantly, given that the observed geometric mean diameter was 141.2 ± 7.7 nm, the enhancement is unlikely to be due to primary volcanic particles such as volcanic ash, which typically have sizes of several micrometers. Instead, it was attributed to secondary aerosols formed and grown through photochemical reactions of the large amounts of SO2 emitted from the volcano (Boulon et al., 2011; Twigg et al., 2016).

Figure 7a shows that NCN were markedly low in the Indian Ocean (148 ± 229 cm-3) and increased slightly in the order of the East China Sea, the South China Sea, and the Strait of Malacca. The volcanic-influence period exhibited the highest concentration of 2834 ± 1330 cm-3. The geometric mean diameter showed a pattern similar to that of NCN (Fig. 7b): it was the smallest in the Indian Ocean (65.9 ± 19.9 nm) and the largest during the volcanic-influence period (141.2 ± 7.7 nm). In the Indian Ocean, all three modes exhibited the lowest number concentrations among all sea areas. During the volcanic-influence period, the Aitken mode had the second-lowest number concentration (339 ± 124 cm-3, higher only than the Indian Ocean), whereas the accumulation mode showed an exceptionally high concentration (2339 ± 1127 cm-3). Given that approximately 80 % of the total aerosols were composed of accumulation mode particles, this finding is consistent with the geometric mean diameter results.

Meanwhile, the East China Sea, the South China Sea, and the Strait of Malacca had broadly comparable geometric mean diameters, differing by less than 10 nm. However, the number concentrations of each mode showed distinct differences among the three seas (Fig. 7c–e). The nucleation mode was the lowest in the South China Sea and the highest in the Strait of Malacca. In contrast, Aitken and accumulation modes were the lowest in the East China Sea, and although they were the highest in the Strait of Malacca, their differences relative to the South China Sea were minor.

Figure 8a and b present that NCCN were the lowest in the Indian Ocean (30 ± 34 cm-3 at 0.2 % SS and 67 ± 60 cm-3 at 0.6 % SS) and the highest in the South China Sea during the volcanic-influence period (1460 ± 473 cm-3 at 0.2 % SS and 1735 ± 469 cm-3 at 0.6 % SS). However, the NCCN/NCN ratios were the second highest in the Indian Ocean (Fig. 8c and d). In particular, the difference in the ratios at 0.6 % SS was considerably smaller than the difference in NCCN between the two regions. This suggests that, despite the smallest aerosol sizes in the Indian Ocean, a substantial fraction of the aerosols consisted of hygroscopic components and thus effectively acted as CCN. Meanwhile, the NCCN/NCN ratios at 0.2 % SS during the volcanic-influence period were exceptionally high, which was likely due to the contribution of hygroscopic sulfate aerosols in addition to the largest aerosol sizes.

A notable feature emerged when comparing the South China Sea and the Strait of Malacca. Although NCN were slightly higher in the Strait of Malacca (1226 ± 517 cm-3 in the South China Sea vs. 1449 ± 1050 cm-3 in the Strait of Malacca), NCCN were actually slightly higher in the South China Sea (233 ± 47 cm-3 at 0.2 % SS and 601 ± 94 cm-3 at 0.6 % SS) than in the Strait of Malacca (225 ± 85 cm-3 at 0.2 % SS and 536 ± 353 cm-3 at 0.6 % SS). The NCCN/NCN ratios at both supersaturations were likewise higher in the South China Sea than in the Strait of Malacca. However, when accounting for measurement uncertainties, the ranges of both ratios at 0.2 % SS overlap slightly, whereas at 0.6 % SS, they remain distinct without overlap (Table S2 in the Supplement). Nevertheless, considering the similarity in the geometric mean diameter between the two regions (Fig. 7b), the difference in the NCCN/NCN ratios could be primarily governed by κ. This suggests that aerosols in the South China Sea were more hygroscopic and therefore more readily activated as CCN than those in the Strait of Malacca. The major source of aerosols in the Strait of Malacca is the substantial ship emissions emitted from heavy cargo traffic (Saputra et al., 2013; Geng et al., 2023). Given that such emissions are typically hydrophobic, it is reasonable to expect lower κ values in the Strait of Malacca than in the South China Sea.

NCCN observed in this study were broadly consistent with previous studies. In the East and South China Seas, NCCN has been reported to range from several hundred per cubic centimeter in remote areas (Atwood et al., 2017; Gong et al., 2023) to several thousand under continental influence (Gao et al., 2020; Ou et al., 2025). Those ranges were comparable to the results of this study (Figs. 8a, b and S8a in the Supplement). In contrast, in the Indian Ocean, which is considered a background environment, Dournaux et al. (2025) reported NCCN in the range of tens to a few hundred per cubic centimeter. These values were consistent not only with the present study but also with observations from other pristine marine regions (Park et al., 2020; Sanchez et al., 2021; Tatzelt et al., 2022).

Figure 9 shows the average aerosol size distribution for each sea area, along with the fitting results for each mode and their total sum. The fitting parameters for each mode and residuals are summarized in Table S3 in the Supplement. The nucleation mode was carefully interpreted, as it was fitted within a significantly restricted geometric mean diameter range (15–25 nm) to account for occasional voltage instabilities in the smallest size bins of the SMPS.

Submicron aerosols dominated the total number concentrations across all regions. In contrast, while the contribution of the coarse mode to the number concentrations was minor, these particles could play a substantial role in CCN activation at low supersaturations.

In the East China Sea, South China Sea, and Strait of Malacca, three submicron modes were observed, but the distributions were characterized by prominent Aitken and accumulation modes. Although the relative number concentrations and contributions of both modes varied somewhat among the sea areas, these distributions appeared to reveal the typical characteristics of anthropogenic influences and secondary aerosol formation (Bates et al., 2004; Seinfeld and Pandis, 2016; Ueda et al., 2016). Notably, the Aitken mode in the Strait of Malacca showed the largest geometric standard deviation of 1.98 among all fitting results, suggesting complex mixing of marine background aerosols, continental aerosols, and ship exhaust. Consequently, given that the majority of the accumulation mode particles can act as CCN, NCCN in these three regions are likely to be highly sensitive to the actual size and chemical composition of the Aitken mode, which accounts for a substantial fraction of the aerosol population.

In contrast, the Indian Ocean exhibited much lower number concentrations across all three submicron modes, but a relatively distinct separation among these modes. Especially, the accumulation mode made the most significant contribution (approximately half of the total), indicating that although NCN in the Indian Ocean were significantly low, a substantial fraction of the particles could act as CCN. However, according to previous studies (Kompalli et al., 2020; Dournaux et al., 2025), even in the Indian Ocean, the dominant aerosol mode can vary by region and air mass origin, with the Aitken mode sometimes prevailing and the nucleation mode not always being observed.

Finally, based on the residuals, the performance of the fitting for the volcanic-influence period was relatively poor compared with the other four regions (Table S3). Although four submicron peaks were identified in the measured size distribution (<40, ∼70, ∼170, and ∼400 nm), increasing the number of submicron modes from two to three reduced the RMSE by less than 10 %. Consequently, only the nucleation and accumulation modes were ultimately fitted, resulting in substantial underestimation at around 50 and 600 nm (Fig. S9 in the Supplement). Nevertheless, a distinct accumulation mode (Nt=2468 cm-3, Dpg=165.9 nm, and σg=1.43) was observed during the volcanic-influence period. This is because most aerosols during this period were likely sulfate-dominated particles (i.e., non-sea-salt sulfate aerosols), which originated from the oxidation of SO2 emitted by the Taal volcano and the subsequent condensation of the resulting H2SO4.

Overall, these results provide direct evidence that applying a single, uniform parameterization of marine aerosols in models is highly problematic and that at least regional differentiation is required to represent them. Moreover, these differences in aerosol size distributions across sea areas can lead to substantial variations in CCN efficiency for a given supersaturation condition, as evaluated in terms of κ and CCN spectra in the following section.

Figure 10 presents the mean and uncertainty ranges of the DC and κ for each supersaturation for each sea area. Note that at lower supersaturations, the uncertainties in CCN measurements are greater. Moreover, since SMPS measures NCN in discrete size bins, DC can fluctuate significantly with only a slight change in NCCN when NCN are very low. The combined effects of these factors can result in considerable uncertainties at low supersaturation (i.e., 0.2 % SS), especially where NCN are very low (e.g., the Indian Ocean). Therefore, our interpretation focused primarily on trends and relative comparisons.

According to Köhler theory, higher supersaturations allow progressively smaller particles to activate; consequently, DC is smaller for higher supersaturation. As an exception, during the volcanic-influence period, similar DC were observed across the 0.4 %–1.0 % SS due to the very narrow distribution of the accumulation mode. Even at the same supersaturation, substantial differences in DC were evident among sea areas. Larger DC – such as those observed in the Strait of Malacca – indicate that only relatively large particles could act as CCN because they had low κ.

Interestingly, κ showed a decreasing trend with increasing supersaturation and spanned a wide range of 0.05–0.3, reflecting variations in aerosol composition and mixing state across sea areas and aerosol sizes. Higher κ values were observed in the Indian Ocean and at certain supersaturations in the South China Sea. This implies that aerosols of the same size in the two sea areas were more likely to activate as CCN at a given supersaturation compared to the other sea areas that showed lower κ values. These supersaturation-dependent DC–κ characteristics influence aerosol activation into CCN and ultimately shape patterns of CCN spectra.

The significantly low κ in the East China Sea was also reported by Gong et al. (2023). Rather than reflecting the extremely low intrinsic κ of aerosols, this was likely due to activation competition under high NCN conditions with limited water vapor availability, highlighting a limitation of the CCN-derived κ method. Although the κ calculation methods differ, the κ in the South China Sea reported in previous studies (Atwood et al., 2017; Ou et al., 2025) were higher than those obtained in this study. According to Nair et al. (2024) and Dournaux et al. (2025), κ values at 0.4 % SS obtained in the Equatorial and Southern Indian Ocean were similar to those derived in this study across the Indian Ocean.

Figure 11 shows measured mean CCN spectra along with fittings by two- and three-parameter formulae. The Twomey formula NCCN=C×SSα (Twomey, 1959), where C and α are the fitting parameters, is the two-parameter formula. A three-parameter formula NCCN=N×(1-exp⁡(-B×SSβ)) was proposed by Ji and Shaw (1998), where N, B, and β are the fitting parameters. These parameters and coefficients of determination for these fittings are summarized in Table 2.

When applying the Twomey formula, relatively low coefficients of determination were obtained for the South China Sea and the volcanic-influence period (0.783 and 0.710, respectively). In contrast, the Ji and Shaw formula yields coefficients of determination close to unity across all regions. Although this improvement was expected, given the inclusion of an additional parameter, it highlights the limitations of the traditional Twomey formula. In particular, the Twomey formula tends to overestimate NCCN at low supersaturations representative of real atmospheric conditions, indicating that its application in cloud modeling requires careful consideration.

Meanwhile, as indicated by the fitted curves, the Ji and Shaw formula predicts that NCCN do not increase infinitely with increasing supersaturation but instead asymptotically converge toward the parameter N closely related to NCN and size distributions. Consequently, in regions with abundant CCN, such as the South China Sea and Strait of Malacca, the influence of aerosol variability on CCN is limited. In contrast, in regions with very low NCCN, such as the Indian Ocean, cloud properties can change significantly as they respond sensitively to aerosol variability.

Parameter β reflects the sensitivity of NCCN to supersaturation. β is directly related to DC. Larger β indicates that NCCN responds more strongly to supersaturation variations. The effect associated with differences in β was especially evident when comparing the South China Sea and the Strait of Malacca. Although the two sea areas exhibited similar N values, β was the highest in the South China Sea and the second lowest in the Strait of Malacca. As a result, NCCN differed substantially between the two sea areas at 0.4 % and 0.6 % SS. This contrast could be attributed to the smaller and narrowly distributed DC and the higher κ in the South China Sea than in the Strait of Malacca, leading to stronger supersaturation-dependent aerosol activation. These differences in β suggest that variations in updraft velocity or subtle changes in thermodynamic conditions can induce substantial differences in cloud microphysics.

In the Indian Ocean, mean CCN spectra of this study are similar to some of the aircraft CCN measurements during the 1999 INDOEX campaign (Hudson and Yum, 2002). These were classified as “Clean” at low altitudes south of the ITCZ over the Indian Ocean. Considering the gap of more than 20 years between these two observation periods, this consistent result provides strong evidence that the Indian Ocean is a pristine remote area.